How to Audit Your Shopify Store’s AI Visibility

A real audit answers four questions: which AI engines cite your store on the queries that matter, why they cite the brands they do instead, what’s structurally broken on your side, and which actions would move your mention rate. The DIY version of this gets you the first two questions in an afternoon. The other two are the work.

The 7 steps

Step 1 - Pick 5 priority queries in your category

Don’t optimize for everything. Pick five buyer queries where citation matters most for revenue. Spread the prompt structure, not just the keywords:

• One top-of-funnel best-in-category (“best [category] for [use case]“)
• One negative or comparative (“worst [category]” or ”[your biggest competitor] alternatives”)
• One budget or spec (”[category] under $X” or ”[category] for [specific scenario]“)
• One forum-pull (“what [category] do people on Reddit actually buy”)
• One specific to your differentiator (the product attribute you want the model to associate with your brand)

Different prompt shapes pull different citation sets from the same model. Keyword variations matter less than structural variations.

Step 2 - Run the 5 prompts across 5 engines, fresh session each

ChatGPT (no memory, fresh session), Google AI Mode, Google AI Overviews, Gemini, Copilot. 25 queries total. Screenshot every answer.

What to capture per answer:

• The brands cited, in order
• The third-party publications cited, in order
• Your brand’s position if mentioned
• Your closest competitor’s position

Step 3 - Inspect the citation pattern

Five prompts × five engines is enough to see the pattern. You’ll typically find:

• Two or three publications the engines extract from across most of your category prompts. Those are your priority citation targets, the third-party sources the model trusts in your category.
• A small cluster of brands that show up in 80%+ of answers regardless of prompt phrasing. If you’re not in that cluster, that’s your gap.
• At least one prompt where the engine recommends the wrong brand for the use case. That’s your wedge, the query where the model has incomplete information you can fill.

Step 4 - Audit your PDPs against the questions the prompts asked

Pick the top 3 PDPs an ad campaign would route spend to. For each:

• Lead-with-the-answer test. Does the first 200 words of the page actually answer the prompt the buyer is asking? Not “here’s our brand story.” The answer to the question.
• Schema validity. Run the page through Google’s Rich Results Test or Schema.org validator. Look for Product, Offer, AggregateRating, BreadcrumbList. Broken schema makes the page unreasonable for the model to cite.
• Freshness signal. Last meaningful content update on the PDP. If it’s been 18 months, AI engines are deprioritizing relative to fresher pages in your category.
• Citation footprint. Search the product name plus the priority publications you identified in Step 3. Is your product reviewed there?

If you’d rather skip the manual check on a product page, proproductpage.com runs the same lead-with-the-answer, schema, freshness, and citation-footprint checks on any product URL and ships a rebuilt page back. Free single-product check, paid build from $6.

Step 5 - Audit Shopify configuration

Three settings to verify:

• Settings → Sales channels → Agentic Storefronts → Preview products → Catalog Mapping. Every product description should be present. Orange warnings on missing fields are blocking, the model can’t reason from a blank description.
• robots.txt customization. Shopify’s default template names exactly seven user-agents, *, adsbot-google, Nutch, AhrefsBot, AhrefsSiteAudit, MJ12bot, Pinterest (verified separately in AI Visibility vs Traditional SEO for Shopify Stores →). AI crawlers (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, GoogleOther, Google-Extended) fall under the wildcard, never named. If you’ve customized your robots.txt.liquid to disallow paths, verify those disallows aren’t accidentally blocking AI crawlers from product or collection URLs.
• Checkout configuration for ChatGPT-attributed traffic. If you’re running or planning to run ChatGPT Ads, the checkout flow needs the attribution params propagated through cart-to-thank-you. A server-side Conversions API + JS Pixel dual-shipped with a shared event_id is the configuration that produces clean attribution.

Step 6 - Score yourself against the CRS rubric

Your Contextual Relevance Score, the number that decides what every dollar of ChatGPT Ad spend is worth, is built from five components. Score each 0-10:

• Schema completeness on advertised PDPs (0 = missing, 10 = full Product + Offer + AggregateRating + BreadcrumbList)
• Freshness (0 = no meaningful updates in 18+ months, 10 = monthly meaningful updates)
• Third-party citation footprint in the priority publications from Step 3 (0 = nowhere, 10 = cited in 3+)
• Question-answering structure on the top 3 PDPs (0 = brand story first, 10 = direct answer in first 200 words)
• Robots.txt + Shopify config (0 = blocking AI crawlers + Agentic Storefronts orange warnings, 10 = explicit AI-crawler allowlist + clean Catalog Mapping)

Sum out of 50. Below 25 is the foundation problem, running ads on top of this pays the upper end of the CPC band. Above 35 is competitive, you can run ads without the foundation tax. Between is the work.

Step 7 - Pick the closest action that moves mention rate

You won’t fix everything in week one. Rank your gaps by leverage and pick one:

• Schema fix on a PDP the ads will target, same-week impact once re-crawled.
• Freshness rewrite on a stale priority page, 2-4 weeks until citation pattern shifts.
• A new third-party citation in a priority publication, 1-3 months on editorial timelines, but the durable one.
• Robots.txt allowlist for AI crawlers, same-day effect, but only matters once the rest of the foundation is in place.

Ship the one. Re-run the 5×5 prompt grid in 30 days and see what moved.

Where the DIY ceiling hits

Steps 1-3 are doable in an afternoon and produce real signal. Steps 4-7 are where DIY breaks down at scale:

• Schema validation across the full catalog (not just 3 PDPs) wants tooling, not eyeballing.
• Freshness scoring across hundreds of pages wants automation.
• CRS scoring across four engines wants a multi-engine query harness, not browser tabs and screenshots.
• Citation-gap analysis wants a categorized publication map for your category, not ad-hoc searches.

For background on what changes structurally between SEO and AI visibility: AI Visibility vs Traditional SEO for Shopify Stores →. For the agency-selection layer once you’ve audited and want to act: Best ChatGPT Ads Agency for Shopify Stores: 6 Checks →. For what an agency layer actually does end-to-end: What Is a ChatGPT Ads Agency? →.

FAQ

Can a Shopify operator run this audit themselves? Yes for steps 1-3 (prompt selection, engine queries, citation inspection). That’s an afternoon’s work and produces a real picture of where you stand. Steps 4-7 (schema audit at scale, freshness scoring, robots.txt allowlist, CRS rubric across all four AI engines, citation-gap analysis) are where the DIY ceiling hits. PLUS and PRO retainers bundle CRS recommendations on advertised PDPs as part of the engagement; PRO adds Message-Match Strategy on top.

How many prompts should I run per category? Five at minimum, five engines apiece. Spread the prompt structure: “best [category] for [use case]”, “worst [category] to avoid”, “budget [category] under $X”, “alternatives to [biggest competitor]”, and a question phrased the way buyers actually ask in your category. Different prompt shapes pull different citation sets, keyword variations matter less than structural variations.

What’s the difference between an AI visibility audit and a CRS audit? Same audit, two framings. AI visibility is the surface (which AI engines mention you, in which prompts, with what third-party citations). CRS is the score (the rubric: schema completeness, freshness, third-party citation footprint, question-answering structure, robots.txt + Shopify config). A complete diagnostic produces both: the visibility scoreboard plus the CRS rubric explaining why your number is what it is.

Which engines should I check? ChatGPT is the priority. Run the same prompts against AI Mode, AI Overviews, and Copilot to see where you are or are not appearing across the broader answer surface buyers are using.

---

If you want this signal four times a week, get the Wire.

---

Correction policy: if anything in this post is wrong, we’ll fix it publicly with a date-stamped note. Email corrections to support@paidaisearch.com.